The Ladder Problem
Disclaimer: The views expressed here are my own and do not represent any employer, past or present. External links are provided for convenience; I’m not responsible for their content or subsequent changes.
I was out to lunch the other day with a friend of mine who is a C-something at a large company.
They’re in the early innings of their AI adoption at their company, and like a lot of places, planning.
He wants to begin using some AI agents within his group.
His boss is leaning heavily on him to show ROI, defined as AI-driven efficiency evidenced by headcount reduction.
My friend expects some natural attrition from people leaving, as well as headcount reduction when older employees retire.
He’s intent on arming mid-career operators, potentially the highest-leverage cohort, with AI tools.
He then mentioned that he would need to keep hiring juniors so they could grow into senior roles.
I almost laughed when he said this, not because it was funny, but because what he said alluded to one of those profound things that are often found hiding in plain sight.
Every senior someone was once a junior someone at a company.
AI will soon afford companies the option to dramatically reduce or even eliminate junior hiring.
From a purely existential standpoint, a company that stops hiring juniors is hollowing itself out.
Well, let me reframe that.
Stupid for most companies.*
Let me run an unthoughtful thought experiment for a moment.
On some level, the sum total of a company’s results stems from the myriad decisions made over the course of a minute, a day, a month, a year by its employees.
At the atomic level every employee optimizes for something in their jobs.
Everyone has an objective function.
Why do senior roles seem safer at this moment from an AI takeover?
Well- you’d think it would be because they do so much that they would be hard to “automate” (in the dumbest abstracted use of that term).
But in fact, senior roles often involve ambiguous problem framing.
Senior positions sit on a spectrum between vaguely thinking about things and making irreversible decisions under uncertainty.
You want to be operating on the right side of the spectrum.
AI will enable different levels of leverage for optimization given the role.
For some tasks and even some roles and industries, AI quickly saturates the optimization space and approaches local maxima, and in some cases maybe even approaches global maximums.
You can’t do any better.
For many roles in many companies it’s preferable to pay for an employee that can produce a multiple of their output with the increased leverage and optimization enabled by AI rather than let that employee go.
For a company that has already been disciplined and running lean, AI-enhanced leverage- giving everyone an AI exoskeleton, is many times, more optimal than headcount reductions.
We are going to use the below image as a two-fer:
The ladder represents both the beginning middle and senior levels of your own career.
And the ladder represents the junior, middle, and senior management at a company.
I was a junior trader before I was a senior trader.
I was a junior portfolio manager before I was a senior portfolio manager.
A ladder without lower rungs is not a ladder.
It’s a ceiling.